The risk of contamination by contact with patients with tuberculosis is reduced by isolation of patients until negativation of direct sputum analysis for the research of tuberculosis bacilli. To evaluate the efficacy of this isolation, we compared, in 32 patients with active tuberculosis, the results of direct examination and culture of the sputum and the clinical outcome. Thirty-two successive patients hospitalized in the same internal medicine unit, received antituberculosis drugs and had 3 sputum examinations per week with direct analysis and culture until negativation of the 3 direct examinations. Then, isolation ended. At the time of direct-negativation, 14 of the 32 patients kept positive cultures. In the 18 remaining subjects, the cultures became negative, about seven days before direct-negativation. Patients with negative cultures had more frequently weight increase (83% versus 71%), were more rapidly without fever (11 days versus 19 days), had less cough and had less severe radiologic disease (50% versus 75%) compared to patients with positive cultures but these differences were not statistically significant due to the small sample size. In tuberculosis patients, 3 successive negative direct sputum examinations do not eliminate the risk of tuberculosis transmission, specially to hospitalized or immunocompromised patients. The risk of contamination in these cases, although unknown, may be weak. Terminating isolation should not be based on sputum examination alone, but also on other factors such as the clinical course (resolution of cough and fever, weight), the initial number of bacilli, and the severity of the radiological lesions.